I think the article is somewhat over-representing the difficulty here. Once you're at the team selection screen and choosing your lineup, there are only 15 possible combinations to choose from. Once you factor in that many/most teams are designed around one or two specific synergies, and that your opponent's team is only partially known (you see their Pokemon species but not the moves, stat distributions, etc), which puts huge error bars around whatever prediction you're trying to make, it usually turns out that you're really only picking from 1-3 realistic choices, and there's a very paper-scissors-rock nature to it that you can't really "learn" in the ML sense.
I think you could have gotten equivalent results on such a predictor using much simpler regressions and/or heuristics, once you've already fixed the matchup.
(Also, I just think it's funny how the paper keeps citing "(Zheng, 2020)", etc, like it's a scholarly article or something. Aaron Zheng is a VGC YouTuber and what is being cited is just an online guide a la GameFAQs)
The soft prediction metric seems especially ridiculous to me. If I'm not mistaken, just picking at random gets better results than their ML selection at >= 5 predictions (1-(2/3)*5 > 0.8438).
However:
> your opponent's team is only partially known (you see their Pokemon species but not the moves, stat distributions, etc)
That's not true in the main competitive live format (e.g. NAIC 2025 which is the main case study here). These tournaments are "open team sheet", aka. moves, ability and held items are known (but not IVs/EVs).
I'm not sure whether this is the case on Smogon though, which means they might even be mixing two completely different datasets...
Most of my experience is with pre team preview singles (where there was an entirely different meta of blindly choosing a lead that would match up favorably against the set of other common leads), but my understanding was that VGC has a handful of Pokemon (Smeargle...) with a P_lead/P_bring ratio of 1.
> Once you're at the team selection screen and choosing your lineup, there are only 15 possible combinations to choose from.
Nit: there are 15 possible lineups (i.e. combinations of 2 pokemons to start the battle with) but there are 90 possible teams if you also factor in the other 2 pokemons in the back.
Legit curious what the use case would be, that would justify Apple adding it in. Like, when do you need to text someone who's within Bluetooth range but somehow has no WiFi or cell reception?
When you're at a protest and the government shuts off the internet in response to the protest. It's happening right now in Togo, has been for ten days (https://pulse.internetsociety.org/en/shutdowns/).
My eyes are opened as to how much more power the people would have if cell phones were all mesh network devices, especially as we enter a world where having a working cell phone is easier than having running water or food.
I'm not holding my breath--it would seem that keeping the people down is more profitable.
But if it happens we'll have to figure out how to write partition tolerant apps, which I think would be a lot of fun. It would also make "going viral" so much more apt, as you might catch the content from people you got physically close to.
I dunno, that still sounds like a perfect use case for a third-party app like the one this post is about. I'm not sure government crackdowns are a core enough experience that it needs a first-party app from Apple.
This is admittedly a rare and minor use case, but maybe on a plane if you’re not sitting next to each other? Last time I flew I saw two teenage girls communicating by typing into the same note file and Airdropping it back and forth for hours - it struck me as very silly that there was no messaging interface that they could use instead.
when do you need to text someone who's within Bluetooth range but somehow has no WiFi or cell reception?
There's no cell service or wifi at my neighborhood movie theater. If I could send her a message when she's up, I could tell my wife to bring me back a box of Sno-caps.
I dunno, I bet if it was widespread people could come up with applications. Like, telling people doesn’t necessarily leave a record, so you could talk about tomorrow’s plans and then send a summary text so everyone has a record and all the details.
“Coded PHY” Bluetooth has a range of up to a kilometer! Once you add mesh forwarding, you could probably cover quite some distance on moderately busy hikes.
On top of that “It would be so easy” is almost never true for a billion users network with all kinds of edge cases. Seems like a very narrow use case when there’s things missing from iMessage that could be way more appealing for a bigger group of users.
The technical details are often not the tricky part of new features. You have to integrate it into the existing app that people know and use, explain how it works, maintain it forever etc.
“With iOS xx, now you can message your loved ones even without any cell or Wi-Fi signal from up to a mile away! Simply make sure you already have an iMessage conversation with them started while you still have signal.”
This happens at festivals - despite being largely offline events, you still want to text your friends "hey I'm over at <place>", but the one rural cell that's usually empty for 362 days out of the year is getting DDoSed by 50000 people suddenly arriving one weekend.
Definitely grab them from Standard Ebooks. Not as huge a selection as Gutenberg, obviously, but still several lifetimes worth of reading, and the quality is way higher.
*facepalm* You're correct. I misread it every time, including the first time I scanned your comment. (A truly strange experience since that is not a common occurrence for me. I simply never saw a "v" there.)
It sure looks like the author did his due diligence; he has a chart of all the different phrases in the payload which triggered the 403 and they all corresponded to paths to common UNIX system configuration files.
Nobody could prove that's exactly what's happening without seeing Cloudflare's internal WAF rules, but can you think of any other reasonable explanation? The endpoint is rejecting a PUT who's payload contains exactly /etc/hosts, /etc/passwd, or /etc/ssh/sshd_config, but NOT /etc/password, /etc/ssh, or /etc/h0sts. What else could it be?
Funny seeing this on the front page – I'm coding a project as I'm browsing this that makes heavy use of TikZJax.
Overall, I'm impressed by how seamlessly it works when it does work. But it's not perfect:
- Some core library functions (for example, most types of fill patterns) simply don't work or aren't implemented for some reason.
- There are a few long-standing bugs. For instance, if using the intersections library to compute the intersection of a line and a circle, it straight-up crashes the entire TikZJax process. Intersections of two lines or two circles are fine, but circle+line fails. My attempts at diagnosing this seem to indicate that it's running out of stack space, so maybe the original TikZ code uses some inefficient recursive algorithm to compute this intersection, and this exceeds some stack size limit that the WebAssembly version introduces. I'm not sure and I haven't been able to get much traction.
- The project doesn't seem to get any love from the original developers anymore. I've filed multiple bugs for months now that never get any form of acknowledgement.
- The build process is pretty convoluted and difficult to reproduce (to try to fix those aforementioned bugs myself), which I guess is what you'd expect from a project that attempts to cross-compile a 20-year-old macro package for a 50-year-old Pascal codebase for rendering in the browser.
Overall I'm very glad TikZJax exists and there's still no better-looking and convenient-to-author diagramming language than TikZ itself. But there's definitely rough edges.
I think you could have gotten equivalent results on such a predictor using much simpler regressions and/or heuristics, once you've already fixed the matchup.
(Also, I just think it's funny how the paper keeps citing "(Zheng, 2020)", etc, like it's a scholarly article or something. Aaron Zheng is a VGC YouTuber and what is being cited is just an online guide a la GameFAQs)
reply